• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

Artificial Intelligence thinks mushroom is a pretzel

Mader Levap said:
My point was that these NNs are failing in very different way than human visual system. Human never would see panda as gibbon just because some pixels in picture were very slightly brightened or darkened. Artificial neural networks "perceive" world very differently than humans, get over it.
Click to expand...
You identify the panda easily because you know you are looking for one. You look at a photo and immediately see there is an animal and you ignore the rest of the image. The NN does not know anything. It does not know that a blob of similarly coloured pixels represents an object. It has nothing to guide it, it looks at all the information contained in the pic.
Why did my classifier just mistake a turtle for a rifle?

Here is an interesting article concerning what can be learnt from modeling part of the neocortex.

We developed a mathematical framework to analyze both the structural and the functional topology of the network, integrating local and global descriptions, enabling us to establish a clear relationship between them. We represent a network as a directed graph, with neurons as the vertices and the synaptic connections directed from pre- to postsynaptic neurons as the edges, which can be analyzed using elementary tools from algebraic topology
Click to expand...
The microcircuit, formed by ~8 million connections (edges) between ~31,000 neurons (vertices), was reconstructed from experimental data, guided by biological principles of organization, and iteratively refined until validated against a battery of independent anatomical and physiological data obtained from experiments
Click to expand...
 
The idea is to, as technology allows, first model a rodent brain and eventually a human.
Here is a great talk by Prof. Idan Segev.

The video is crap, but you can hear what he is saying. Start about 10min in if anyone is interested.

 
arthwollipot said:
There are some (I happen to not be one of them) who feel that true General Artificial Intelligence is similarly impossible not because of a lack of technology, but because the laws of the universe prevent such a thing from existing.
Click to expand...

The existence of human brains demonstrates that a machine capable of general intelligence is possible given the laws of physics in our universe.
 
Roboramma said:
The existence of human brains demonstrates that a machine capable of general intelligence is possible given the laws of physics in our universe.
Click to expand...
But if we have to replicate a human brain in an artificial medium, that's going to be a darn hard project. The human brain isn't a machine - it is not manufactured. It is biology. It is born. To those who hold this view (and I repeat, I am not one of them) there is an insurmountable gulf between what human beings are or ever will be capable of manufacturing, and what can grow naturally due to biology.
 
arthwollipot said:
But if we have to replicate a human brain in an artificial medium, that's going to be a darn hard project. The human brain isn't a machine - it is not manufactured. It is biology. It is born. To those who hold this view (and I repeat, I am not one of them) there is an insurmountable gulf between what human beings are or ever will be capable of manufacturing, and what can grow naturally due to biology.
Click to expand...

I don't see any difference of kind between a biological system and a manufactured system. The products of biology and the products of our technology are both systems of atoms interacting according to the laws of physics.

Whatever principles allow a human brain to function as it does are physical principles. I'm sure there are constraints on how such systems can be designed and what sort of manufacturing techniques can be applied. But there is at least one design that we know works and at least one manufacturing technique that we also know works (in both cases, the ones that exist in nature).

Many of the limitations of human brains are constraints imposed by our biology: limitations of energy related to how much food our ancestors could find and consume in a day, limits of space based on the size of the human skull, limits of access to data based on what is available to our senses, networking constraints based on the lack of something analogous to wifi (the closest we have is speech). But all of those are specific to humans and are not general limitations of intelligence. If we learn to build systems that work similarly to human brains we won't have the same energy constraints, space constraints, input constraints, or networking constraints.

The same logic applies to the constraints of the manufacturing process. Our biology is constrained by having to start from a single cell, for instance. To building things out of proteins or things that protiens can build. Of being self-reproducing (so for instance whatever size constraints there are on the organism also exist for the machine that builds the organism, because they are one and the same. That's not true of cars for instance, which while they have a size constraint of having to fit on roads, can build built in factories that don't fit on roads). Etc. Again at least some of the constraints on the manufacturing process of human brains are specific to our biology and don't apply in general to intelligent systems.
 
Last edited:
arthwollipot said:
But if we have to replicate a human brain in an artificial medium, that's going to be a darn hard project. The human brain isn't a machine - it is not manufactured. It is biology. It is born. To those who hold this view (and I repeat, I am not one of them) there is an insurmountable gulf between what human beings are or ever will be capable of manufacturing, and what can grow naturally due to biology.
Click to expand...

The argument is rather moot, when technology can simulate biology. E.g., proteins are produced by biology, but you can download Folding@Home and simulate how they fold. The same applies to neurons. When you understand well enough how one works, you can simulate it.

Just like we can simulate anything else, really. I mean, computers aren't riveted together, but we can simulate a bridge on a computer anyway. Computers don't run on diesel, but we can simulate a diesel engine. Computers aren't liquids flowing through pipes, but we can simulate fluid flows. Etc.

Mind you, just simulating neurons would be a rather inefficient way to do it. But for the purpose of the argument "but computers aren't biological!!!111eleventeen", just the fact that it's possible to simulate biology on a computer is enough to shoot it down.
 
HansMustermann said:
The argument is rather moot, when technology can simulate biology. E.g., proteins are produced by biology, but you can download Folding@Home and simulate how they fold. The same applies to neurons. When you understand well enough how one works, you can simulate it...
Click to expand...
And that's pretty much why I personally do not subscribe to that point of view.

I should also point out here, while I am listing arguments that other people have made, that some people believe that human intelligence can never be achieved by a computer because a computer doesn't have a soul. I assume that we're all in agreement about that particular opinion as well, so having mentioned it I think we can move on.
 
HansMustermann said:
The argument is rather moot, when technology can simulate biology. E.g., proteins are produced by biology, but you can download Folding@Home and simulate how they fold. The same applies to neurons. When you understand well enough how one works, you can simulate it.

Just like we can simulate anything else, really. I mean, computers aren't riveted together, but we can simulate a bridge on a computer anyway. Computers don't run on diesel, but we can simulate a diesel engine. Computers aren't liquids flowing through pipes, but we can simulate fluid flows. Etc.

Mind you, just simulating neurons would be a rather inefficient way to do it. But for the purpose of the argument "but computers aren't biological!!!111eleventeen", just the fact that it's possible to simulate biology on a computer is enough to shoot it down.
Click to expand...

I'm not certain it's sufficient to simulate something. You can't cross a river on a simulated bridge, you can't power a vehicle with a simulated diesel engine, and a simulated fluid can't actually get anything wet. Perhaps intelligence is the same: it can be simulated but that simulation won't be able to do what a real intelligence can do.
 
TragicMonkey said:
I'm not certain it's sufficient to simulate something. You can't cross a river on a simulated bridge, you can't power a vehicle with a simulated diesel engine, and a simulated fluid can't actually get anything wet. Perhaps intelligence is the same: it can be simulated but that simulation won't be able to do what a real intelligence can do.
Click to expand...
I'm not sure what the difference would be between a thing and a 100% accurate simulation of that thing.

You can't cross a real river on a simulated bridge, but you can cross a simulated river. As long as you're doing so in a simulated car.
 
arthwollipot said:
I'm not sure what the difference would be between a thing and a 100% accurate simulation of that thing.

You can't cross a real river on a simulated bridge, but you can cross a simulated river. As long as you're doing so in a simulated car.
Click to expand...

Unless you have a holodeck a simulation is still just a simulation. A simulated intelligence would simulate thought, but not actually think.
 
arthwollipot said:
How could you tell the difference?
Click to expand...

Just because you can't perceive a difference doesn't mean it's not there. Is that cat over there slumbering peacefully or just brilliantly taxidermied? You can't tell but it makes a hell of a difference to the cat.
 
Mate. Thinking is just processing some data. If you get the same data in, and the same result comes out, then how the heck is it any different from thinking?
 
HansMustermann said:
Mate. Thinking is just processing some data. If you get the same data in, and the same result comes out, then how the heck is it any different from thinking?
Click to expand...

If you don't read Chinese and respond to written questions in Chinese characters someone else directs you to write, are you understanding Chinese?
 
You must log in or register to reply here.

Back
Top Bottom